in number of patients. In studies performed in the clinical groups, the effect estimates are usually higher in value but can be low in statistical power due to small sample size. Replica- tion in other samples and joint collaborations with other research centers on (isolated and syndromic) oligodontia would be the best solution to overcome this concern, however it was not possible for the current thesis.
5.3.5 External validity
In studies performed in the Generation R Study, most of the participants were of Dutch ethnicity and belonged to a higher socio-economic class as compared to non-participants. Furthermore, the number of participants with gestational disorders or preterm born children were lower than expected from the population figures in Rotterdam 53. This selection towards
a more affluent and healthy population at baseline may have led to reduced statistical power, due to lower prevalence rates and subsequently affecting the generalizability of our findings to other populations. A representation of more ethnic and socioeconomic subgroups would be necessary to achieve external validity. Thus, meta-analysis and replication of the studied associations in other populations in Europe and other continents would make the generaliz- ability of our findings possible.
5.3.6 Causality
Due to the design of our studies we couldn’t investigate causality of the observed associa- tions. Bradford’s Hill criteria on causation presents the minimal conditions needed to estab- lish a causal relationship between potential exposures and outcomes 59. The studies included
in this thesis fulfilled the Hill’s criteria on causation as presented in supplementary Table S5.1. Randomized controlled trials are often preferred to establish causality and identify mecha- nisms that describe in detail the associations. In order to bring insight on causality, Mendelian randomization can also be applied in the observational studies. For this type of study large sample sizes are needed in order to obtain sufficient statistical power. With the impossibility to obtain similar measurements on dental development from other collaborative cohorts, no causal association could be proven in this thesis. In particular for investigations performed in clinical samples, experiments in animals are the most applicable instrument to detect causal genes in abnormalities of teeth. As dentition of mice is closer to human dentition, mice are mostly chosen to study the genetic complex network of dental development from initial formation 5,6.
5.3.7 Repeated measurements to monitor the developing dentition
Measurements of dental development available for this thesis were cross-sectional, known as one of the least costly epidemiological designs. Because dental development is a continuous and progressive process, beginning at 6th intrauterine week and ending 18 or more years
later, systematic monitoring is necessary not only for scientific research but also for clinical considerations. Taking periodic radiographs from the beginning of the mixed dentition to the complete eruption of the permanent teeth could reveal many developmental problems
and facilitate early treatment intervention 8. Serial DPRs taken at the age 6, 8, 10 and 12 years
would be a necessary step to evaluate and monitor dental development in young patients. Following similar steps as in the clinic would contribute to detect normal variations and disturbances in the general population. Therefore, serial DPRs at the age 6 or 7, 9 or 10 and 12 or 13 years would be ideal steps to measure dental development longitudinally. However considering exposure to X-ray radiation every three years, in cohort studies with a large number of participants, such as the Generation R Study, we suggest at least two time points measurements as necessary to obtain a longitudinal approach of dental development. Mea- surements of dental development in the Generation R Study were available only at the age of 10 years. Due to the lack of radiographic images at earlier ages repeated measurements were not applicable for this thesis. However, we are currently taking DPRs at 13 years old participants in the Generation R Study to achieve the evaluation of dental development as a continuum.
5.3.8 Assessment of dental development
Chronological age is often not a good indicator of the individual’s growth status 60. Therefore
developmental age rather than chronological age can be a useful approach in evaluating a child’s growth status 61. Because tooth development shows less variability than other devel-
opmental components and also low variability in relation to chronological age, dental age can be used with high reliability as a proxy of developmental age 6, 9. Radiological, histologi-
cal and biochemical methods are used to define dental age 62. Histological methods require
extraction or preparation of microscopic sections of at least one tooth from each individual. The biochemical methods are based on the racemization of amino acids and used to estimate the age when the individual died. Thus, these methods are not applicable in living individuals for scientific and ethical reasons 62. Besides, these are quite expensive and require sophisti-
cated laboratory equipment. On the contrary, the radiographic methods are simple, quick, economic, non-invasive and applicable in both individual and population level 63. Among
radiographic methods to estimate dental age, we used Demirjian method to assess dental development due to its advantages and tackled disadvantages as described in Table S5.2.
5.3.8.1 Advantages of Demirjian method
Demirjian method is widely spread in research due its simplicity in application. The ap- proach of the Demrijian method requires the identification of 8 developmental stages in the seven teeth of the left lower quadrant to calculate dental age of and provide an overall proxy of dental development for each subject 64. The characteristics of each developmental
stage are well specified and easily visualized in a radiographic image, assigning two major advantages of Demrijian method, good reproducibility and high intra and inter-examiner reliability 62. Whereas, other radiological methods presented by Schour and Masseler, Nolla
and Moorees require identification of 21, 10, 14 developmental stages, respectively. With the increased number of stages, the detailed visualization of characteristics that represent each developmental stage will be more difficult and less accurate, consequently. As a result of applying these methods, investigators may have to deal often with intra and inter-examiner